[QUOTE=lavalamp;223377]Indeed, 4 flops/cycle, however using that formula still does not yield an answer that matches the reported throughput of the i7 980 XE.

flop/cycle * freq * cores = 4 * 3.33 * 6 = 80 DP GFLOP/s

But the reported performance is 107.55, not 80. 107.55 is what wikipedia says it is with no source, and a lot of other places also say that's what it is, but appear to have copied it straight from the wikipedia page.

If the multiplier was increased from 25 to 27 (as it can be with turbo mode, but only when one core is active), and instead of 4 flops/cycle it is somehow 5, then you can get close to that number:

5 * 3.6 * 6 = 108 GFLOP/s

So it is either an error that was copied and reported as a fact everywhere, or something weird is going on.[/QUOTE]

I trust the talkpage: [url]http://en.wikipedia.org/wiki/Talk:FLOPS#PC_Speed[/url]

[QUOTE=ldesnogu;223382]I don't understand... There's no law, just some much faster CPU/GPU that will make your computer/cluster useless.
[/quote]
In the situation I described, getting a sunroof = getting a GPU, and driving = getting onto the top 5000 list.
If you don't have a sunroof, you can't drive.
If you don't get a GPU, you can't get onto the top 5000 list.

[quote] if your hardware is obsolete use it for some less demanding tasks, and leave all the glory to the big boys with deep pockets.[/QUOTE]
This was supposed to be fun. Distributed computing projects were originally intended to let all computers participate, not just the best ones.

[quote=Oddball;223385]This was supposed to be fun. Distributed computing projects were originally intended to let all computers participate, not just the best ones.[/quote]I've been running obsolete computers for over 20 years now and have built up quite a collection of them.

When GIMPS started a powerful consumer PC was a P90 or so and many 486 machines contributed. At that time, a number of 8-bit machines (BBC micro and Apple ][ for instance) were still in daily productive use. There was no way even a 16-bit 286 machine could contribute to GIMPS and they were still common.

It has never been the case that all computers have been able to participate. Today's equivalent of the 286 is probably a P90 or thereabouts. I retired my PPro less than two years ago; Phil Carmody retired his not too long before. AFAIK, both machines still work fine --- mine certainly does.


Paul

[QUOTE=Oddball;223380]Just out of curiosity, is that true for all the recent AMD processors too? By recent, I mean the K10 series.[/QUOTE]
I think so. Not sure about Phenom, but Phenom II and Athlon II -- definitely.

[QUOTE=Oddball;223385]In the situation I described, getting a sunroof = getting a GPU, and driving = getting onto the top 5000 list.
If you don't have a sunroof, you can't drive.
If you don't get a GPU, you can't get onto the top 5000 list.[/quote]
Driving is a practical necessity and therefore, closer to a right.. Getting on to Top 5000 is a privilege -- you have to earn it.

[QUOTE=Oddball;223385]This was supposed to be fun. Distributed computing projects were originally intended to let all computers participate, not just the best ones.[/QUOTE]
It can still be fun to participate. Unless, by fun, you mean the ability to out-compete others in stats -- in which case, quit whining and get a faster setup.
Secondly, DC projects hardly disallow participants with slower computers. But each project has to decide what timelimit should be imposed on the completion of a work unit. On the one hand we have projects with very loose restrictions (like GIMPS with multi-year timelines or RPS with virtually no restriction other than frequent status updates), to BOINC projects with WU limited to mere hours. The availability of GPU-based codepath will not typically affect this decision, unless and until GPUs become the norm for the project.

PS:- PrimeGrid will still make it harder to get on to Top 5000 -- even without GPUs. What are you going to do about that?

[QUOTE=ldesnogu;223378]What I find funny is that Intel reports much lower numbers: [URL]http://www.intel.com/support/processors/sb/cs-023143.htm[/URL]
Oh wait that's the export compliance metrics page :razz:[/QUOTE]Those numbers are bang on the theoretical peak performance.

Take the i7 965 for example, 4 flops/cycle, 4 cores, 3.2 GHz:
4 * 4 * 3.2 = 51.2 GFLOP/s

[QUOTE=axn;223383]I trust the talkpage: [url]http://en.wikipedia.org/wiki/Talk:FLOPS#PC_Speed[/url][/QUOTE]Yeah, a random guy on there comes out with 79.9 GFLOP/s (because he used a slightly wrong clock speed), so for all intents and purposes he calculated 80 GFLOP/s peak for the i7 980 too.

[quote=Oddball;223372]edit: mdettweiler, I know you mentioned limiting GPUs to certain parts of your project, like megabit primes and primes too small for the top 5000 list. While that's a reasonable idea, it's unlikely that Primegrid would show the same restraint.[/quote]
I'm not so sure about that. We (the NPLB admins) do have somewhat of a rapport with the PrimeGrid administration, so I'm sure we could discuss a general plan between us that could keep GPUs from doing too much damage. Also, keep in mind: PrimeGrid does primarily Proth numbers (k*2^n+1), whereas NPLB does Riesel numbers only (k*2^n-1). If the GPU application can only do LLR tests (Riesel numbers), then in fact it would be quite helpful in allowing NPLB to keep up with PrimeGrid's "big iron" on the Proth side more easily. As it is now we can keep our low-top-5000 drive up with the rising cutoff pretty well, but only by one of our admins keeping a large portion of his farm on it semi-permanently; having some GPUs to maintain that drive would allow us to use the CPUs more flexibly elsewhere.

[quote]
The availability of GPU-based codepath will not typically affect this decision, unless and until GPUs become the norm for the project.
[/quote]
It's only a matter of time, I suppose :sad:

[quote]
PS:- PrimeGrid will still make it harder to get on to Top 5000 -- even without GPUs. What are you going to do about that?[/QUOTE]
Not much. But I wouldn't want to imagine what a Primegrid equipped with GPUs would be like. What would the minimum entry level of the top 5000 be? 1 million digits? Fighting off a pit bull is quite hard, but you'll probably make it out alive. Now if you had to fight off a whole pack of wolves instead...

So take a look at [url=http://primes.utm.edu/top20/trends.php]here[/url].

About 2 years ago PrimeGrid searched for constant-n primes of the form k*2^n-1 with n=333333.
As you can see the 5000th place (red line) was horizontal, which means the entry was about a year on the same digit-level.
PrimeGrid found about 1500 primes with that n and pushed there own 'smaller' primes out of the Top5000!

Hi,

llrpsrc.zip on ubuntu

k*2^n+1:correct result
[QUOTE]
1*2^216091-1 is prime! Time : 113.888 sec.
55185*2^8092+1 is prime! Time : 334.000 ms.
93*2^135908+1 is prime! Time : 69.017 sec.
[/QUOTE]

k*2^n-1:abort

[QUOTE=msft;223431]
k*2^n+1:correct result
[/QUOTE]
This will make many people very happy! Can't wait to try it. :smile:

[QUOTE=frmky;223432]This will make many people very happy! Can't wait to try it. :smile:[/QUOTE]

It is no change,Original source.:cry:

Llr is very complex source.